| With the progress of hardware and software,the application of computer technology is also developing rapidly.The combination of computer technology and traditional cars has given birth to driverless technology,which includes environmental perception technology,positioning technology,intelligent decision-making technology,path planning technology,and control technology,of which environmental perception is an important foundation for the subsequent technology,and the effect of perception determines the driving safety of driverless cars.Lidar is one of the key sensors in the perception technology,and its advantages are that it can measure the distance,the result is more accurate,and it is not easy to be affected by light,but the price of high-resolution Lidar in the market is very high,and the point cloud data collected by low-resolution Lidar is sparse,which seriously affects the judgment of positioning and path decision in the unmanned driving technology.Therefore,this paper proposes a method of aligning multi-lidar point cloud data based on simulated annealing algorithm with optimized genetic algorithm to address the problems related to lidar in current perception technology,and after actual data collection and related experimental analysis,the method has application feasibility in improving lidar alignment accuracy.The main works are as follows.（1）For the problems of noise,spurious points and huge redundant data in the point cloud data collected by Lidar,this paper proposes an adaptive voxel filtering method based on the voxel filtering method,which adaptively adjusts the voxel grid edge length according to the filtering ratio to solve the problems of too few remaining points and loss of key features in the filtering process caused by too large edge length,and the problem of low filtering efficiency caused by too small edge length.（2）For the multi-Lidar point cloud alignment problem,this paper optimizes the genetic algorithm based on the simulated annealing algorithm,combining the simulated annealing algorithm with the genetic algorithm that has a strong ability to search for the local optimal solution and the genetic algorithm that has a strong ability to search for the global optimal solution,which is applied to find the optimal solution of rigid body transformation in the point cloud alignment,so as to combine the multiLidar point clouds and compensate for the sparse single-Lidar point clouds.（3）In order to verify the adaptive voxel filtering method proposed in this paper,the point cloud data are collected in a real scene for filtering.The experiments show that the filtering method in this paper can adaptively adjust the voxel grid edge length according to the required filtering ratio on the basis of retaining the key features of the source point cloud,avoiding the problems arising from the artificially given edge length.（4）To verify the practicality and accuracy of the proposed multi-Lidar point cloud alignment algorithm,multiple 16-line Lidar point clouds are aligned with the reference point cloud on the basis of the actual collected point cloud data,and the alignment results are compared with the results of the traditional point cloud alignment algorithm.The experimental results show that the accuracy of the proposed alignment algorithm is improved compared with the traditional point cloud alignment algorithm,and it is feasible to apply to the problem of finding the optimal solution of rigid body transformation parameters for point cloud alignment. |
